Journal of Computer Science

Original Research Paper

Predicting Smartphone Addiction in Teenagers: An

Integrative Model Incorporating Machine Learning and Big
Five Personality Traits
Jacobo Osorio, Marko Figueroa and Lenis Wong

Department of Information Systems Engineering Program, Peruvian University of Applied Sciences, Peru

Article history Abstract: Smartphone addiction has emerged as a growing concern in

Received: 14-07-2023 society, particularly among teenagers, due to its potential negative impact on
Revised: 20-09-2023 physical, emotional social well-being. The excessive use of smartphones has
Accepted: 18-10-2023 consistently shown associations with negative outcomes, highlighting a
Corresponding Author:
strong dependence on these devices, which often leads to detrimental effects
Lenis Wong on mental health, including heightened levels of anxiety, distress, stress
Department of Information depression. This psychological burden can further result in the neglect of
Systems Engineering Program, daily activities as individuals become increasingly engrossed in seeking
Peruvian University of Applied pleasure through their smartphones. The aim of this study is to develop a
Sciences, Peru predictive model utilizing machine learning techniques to identify
Email: lwongpuni@gmail.com smartphone addiction based on the "Big Five Personality Traits (BFPT)".
The model was developed by following five out of the six phases of the
"Cross Industry Standard Process for Data Mining (CRISP-DM)"
methodology, namely "business understanding," "data understanding," "data
preparation," "modeling," and "evaluation." To construct the database, data
was collected from a school using the Big Five Inventory (BFI) and the
Smartphone Addiction Scale (SAS) questionnaires. Subsequently, four
algorithms (DT, RF, XGB LG) were employed the correlation between the
personality traits and addiction was examined. The analysis revealed a
relationship between the traits of neuroticism and conscientiousness with
smartphone addiction. The results demonstrated that the RF algorithm
achieved an accuracy of 89.7%, a precision of 87.3% the highest AUC value
on the ROC curve. These findings highlight the effectiveness of the proposed
model in accurately predicting smartphone addiction among adolescents.

Keywords: Smartphone Addiction, Machine Learning, Predictive Model,

Big Five Personality Traits, Random Forest

Introduction substance, activity or relationship; they are a set of signs

and symptoms that are influenced by biological, genetic,
In contemporary times, it is widely acknowledged that psychological and social factors” (Castillo-Viera et al.,
people have become increasingly reliant on smartphones. 2022). Considering this definition, smartphone addiction
Despite the availability of studies highlighting this issue, no can be regarded as a psychological addiction due to its
significant preventive measures or actions have been taken repetitive involvement of pleasurable behaviors, leading to
to address this growing problem. It is important to emphasize a loss of control that hinders individuals in their daily
that smartphone addiction has been linked to physical (Li et al., activities (Minsa, 2021). It has been observed that 83% of
2021) and emotional effects (Chen et al., 2021; Cheng and Peru's urban population uses smartphones, with over 50%
Meng, 2021; Lei et al., 2020; Wickord and Quaiser-Pohl, using them for entertainment purposes (Ipsos, 2021).
2022), contributing to psychological challenges in Although smartphones were initially designed for the
individuals’ lives (Lei et al., 2020; Rho et al., 2019). primary purpose of communication, they have become
“Addictions are physical and psycho-emotional widely used for entertainment purposes. This includes
diseases that create a dependency on or need for a activities such as using social networks, streaming series

© 2024 Jacobo Osorio, Marko Figueroa and Lenis Wong. This open-access article is distributed under a Creative Commons
Attribution (CC-BY) 4.0 license.
Jacobo Osorio et al. / Journal of Computer Science 2024, 20 (2): 181.190
DOI: 10.3844/jcssp.2024.181.190

or movies even online shopping, which reinforce factors On the other hand, adults are prone to experience
that contribute to addiction (Nida, 2022). emotional dependence on the device, where they feel the
Besides multiple factors leading to addictive need to be constantly connected due to work and personal
behaviors, it is important to underscore the association responsibilities. Because of this, adults face difficulties in
between addiction and the neuroticism trait (Lei et al., balancing the time they dedicate to it despite what was
2020; Müller et al., 2021). Individuals with a high score previously described, adults have greater self-control
in this personality trait are more susceptible to when using a smartphone, which shows an inversely
distractibility and engaging in obsessive thoughts proportional relationship between age and smartphone
associated with addiction, anxiety, or stress. This addiction (Marengo et al., 2020).
predisposition can lead to the development of addictive Personality Traits
behaviors and a strong dependence on their devices (Li et al.,
2022), causing them to neglect their daily activities in The relationship between the Big Five Personality
Traits (BFPT) and smartphone addiction was analyzed to
pursuit of solitude (Abu-Taieh et al., 2022; Chen et al.,
identify the primary characteristics associated with this
2021), having side effects such as anxiety (Li et al., 2022;
addiction based on individuals' personality traits.
Müller et al., 2021), distress (Chen et al., 2021; Lei et al.,
Subsequently, the relation between each trait and
2020) and stress (Müller et al., 2021). The effects extend
addiction was examined, encompassing the following five
to the individual's mental, physical emotional well-being. traits from the BFPT: Neuroticism, conscientiousness,
It is essential to note that these psychological issues are extraversion, agreeableness openness.
also recognized as contributing factors to addiction Regarding the influence of personality traits, it was
(Cheng and Meng, 2021; Lei et al., 2020). found that extraversion is related to smartphone addiction
To address this issue, various studies have emerged (Toyama and Hayashi, 2022). It has been identified that
emphasizing the necessity of implementing strategies to people with high levels have the need to constantly
mitigate smartphone addiction. It is highly recommended interact and seek out satisfactory social experiences,
to implement programs aimed at raising awareness and leading to the device serving to fulfill their socialization
promoting responsible mobile device usage, both within needs (Eichenberg et al., 2021; Peltonen et al., 2020).
educational settings and within families. However, Moreover, neuroticism is considered the most
despite the current research on the correlation between influential trait (Müller et al., 2021). High levels are
smartphone addiction and personality traits, the utilization associated with a tendency to experience negative
of machine learning algorithms as a solution remains emotions (Müller et al., 2021) and difficulty in managing
largely unexplored. Thus, there is a pressing need to stress or anxiety, making individuals more susceptible to
developing addiction. For this reason, using the device
develop a novel approach that leverages new technologies,
provides a sense of security and control, which leads to an
enabling more accurate prediction of smartphone addiction
increase in its constant use (Lei et al., 2020; Müller et al.,
among adolescents. 2021; Zeighami et al., 2021).
Therefore, this article presents a smartphone addiction Conscientiousness has an inverse relationship with
prediction model that combines machine learning techniques
addiction, whereby individuals with lower levels tend to
and the big five personality traits. The random forest
be less aware of the amount of time they spend using their
algorithm is employed to analyze the relationship between
devices. They may exhibit disorganized tendencies and
the five personality traits and smartphone addiction,
utilizing the "Big Five Inventory (BFI)" and "Smartphone often prioritize less important activities over more
Addiction Scale (SAS)" questionnaires. significant ones (Marengo et al., 2020; Müller et al., 2021).
On the other hand, the traits of openness and
Age Groups agreeableness have not shown a direct relationship
It is important to consider age as it plays a significant with smartphone addiction (Eichenberg et al., 2021;
factor in the relationship between individuals and their Erdem and Uzun, 2022; Müller et al., 2021; Peltonen et al.,
smartphones (Eichenberg et al., 2021). 2020). Although these traits influence other aspects of
On the one hand, considering age as a relevant factor, human behavior, they are not directly linked to
teenagers (Abu-Taieh et al., 2022; Duan et al., 2021) are smartphone addiction.
more susceptible to present smartphone addiction due to In conclusion, there is a direct relationship between the
increased exposure to technology, which begins at an trait of neuroticism and an inverse relationship with the
early age when there is a lack of self-regulation. As a trait of conscientiousness (Peterka-Bonetta et al., 2019;
result, teenagers may have a higher likelihood of being Toyama and Hayashi, 2022) and, despite the existence of
exposed to smartphones and being drawn towards evidence based on the analysis carried out, it is important
engaging with satisfying stimuli, thereby increasing their to highlight that these statements are only hypotheses
vulnerability to smartphone addiction. since addiction is a complex phenomenon that is influenced

182
Jacobo Osorio et al. / Journal of Computer Science 2024, 20 (2): 181.190
DOI: 10.3844/jcssp.2024.181.190

by multiple factors such as the social environment, useful when attempting to understand the influence of
behavior patterns other psychological aspects. each trait or variable in predicting behavior.
Machine Learning Algorithms
Materials and Methods
Machine Learning algorithms are employed to analyze
the collected data and establish a correlation between Cross Industry Standard Process for Data Mining
personality traits and smartphone usage patterns, as (CRISP-DM) is an approach that is structured for the
depicted in Table 1. The aim of this comparison is to development of data mining projects and the creation of
identify the algorithm that best suits the research prediction models (Blasi and Alsuwaiket, 2020). It
objectives and facilitates the development of an accurate consists of six main phases that can be seen in Fig. 1. The
detection model to enhance our understanding of phases are (1) Business understanding, (2) Data
smartphone addiction. Among the algorithms utilized for understanding, (3) Data preparation, (4) Modeling, (5)
predicting behaviors and psychological disorders, we Evaluation (6) Deployment. For the development of the
consider Decision Tree (DT), Random Forest (RF), XG research, the first five phases were carried out.
Boost (XBG) Logistic Regression (LR).
The DT algorithm is known for its interpretability (Lee and
Kim, 2021), enabling us to identify direct relationships
between personality traits and addiction. It delivers
outstanding results in terms of precision and specificity
when predicting behaviors (Makino et al., 2021).
However, while the DT algorithm is effective for behavior
prediction, the RF algorithm outperforms it in terms of
performance (Razavi et al., 2020).
By leveraging an ensemble of multiple decision trees,
the Random Forest algorithm allows for more
comprehensive study and analysis (Abu-Taieh et al., 2022).
The RF algorithm demonstrated superior performance in
detecting mental health crises (Garriga et al., 2022; Xia et al.,
2022) and has demonstrated superior performance in most
cases (Lee and Kim, 2021) due to its ability to identify
complex patterns in the data.
Similarly, the LR algorithm is employed in the
prediction of behaviors. Unlike the others, it is a binary
classification algorithm that estimates the probabilities of
belonging to a specific class, which makes it particularly Fig. 1: CRIPS-DM methodology (Peralta, 2014)
Table 1: Comparison of algorithms for behavior prediction
Reference Algorithm Purpose Complexity Robustness
Chen et al. (2022); Duan et al. Decision Tree (DT) Perform classifications Low complexity, Moderate robustness
(2021); Garriga et al. (2022); and predictions with an easy-to- against noise and
Haque et al. (2021); Kim et al. understand outlier data, as it
(2021); Lee and Kim (2021); structure can be influenced
Makino et al. (2021); Xia et al. by the presence of
(2022) extreme values
Abu-Taieh et al. (2022); Random Forest (RF) Perform classifications Medium complexity, Robustness against
Chen et al. (2022); Haque et al. and predictions with manageable noisy and outlier
(2021); Kim et al. (2021); Lee interpretability data due to its
and Kim (2021); Peltonen et al. combination of
(2020); Razavi et al. (2020); multiple decision
Xia et al. (2022) trees
Chen et al. (2022); Garriga et al. XG Boost (XGB) Enhance the performance Highly complex, Robustness against
(2022); Haque et al. (2021); and accuracy of machine requires more noise and outlier
Lee and Kim (2021); learning models advanced knowledge data thanks to its
Xia et al. (2022) ability to handle
errors
and learn from them
Chen et al. (2022); Garriga et al. Logistic Regression Classification involves Low complexity, Moderate robustness
(2022); Razavi et al. (2020); (LR) estimating the probabilities easy to understand against noisy and
Xia et al. (2022) of belonging to a class and implement outlier data, thereby
or not influenced by the
presence of extreme
values

183
Jacobo Osorio et al. / Journal of Computer Science 2024, 20 (2): 181.190
DOI: 10.3844/jcssp.2024.181.190

Business Understanding related to video games (Peltonen et al., 2020). Finally, the
openness trait is not related to smartphone addiction
With the objective of identifying the personality traits (Wickord and Quaiser-Pohl, 2022).
that are related to smartphone addiction in adolescents
between 12 and 17, a review of the literature and the social Data Preparation
context has been carried out to analyze the personality
This phase consists of data selection, data cleaning,
traits related to smartphone addiction.
data construction, data integration data formatting. To
Data Understanding collect essential data, the Big Five Inventory (BFI)
survey was used, which is designed to evaluate the
The variables used for the prediction model are based personality traits of the participants (Cha and Seo, 2018).
on studies that analyze the big five personality traits In addition to assessing personality traits, our objective
(Eichenberg et al., 2021; Erdem and Uzun, 2022; was to determine the level of smartphone addiction
Peltonen et al., 2020; Wickord and Quaiser-Pohl, 2022) are
among each teenager. Hence, following the BFI
shown as a direct relationship with smartphone addiction
questions, we included the Smartphone Addiction Scale-
this is observed in Table 2. After establishing the variables
Short Version (SAS-SV) survey.
and understanding their relationship with the addiction, a
A study was conducted at a private school in San
connection with avoidance is identified. On the one hand,
Miguel, Lima, Peru, where 118 anonymous surveys were
people who present high levels of neuroticism tend to use
collected from students aged 12-16 years. Each survey
their smartphones as a way of distraction from their
worries because of social anxiety; this could be a reason consisted of 54 closed-ended questions regarding personality
for their preference to keep in touch through social traits and smartphone usage. The responses were rated on a
networks, which leads to an increase in smartphone use. Likert scale ranging from 1-5 for the Big Five Inventory
On the other hand, the trait that is inversely related to (BFI) which has 5 dimensions and is characterized by its
addiction is conscientiousness, since people with lower acceptable internal consistency (Cronbach’s  = 0.75 for
levels are less prone to set limits regarding the use of the neuroticism, 0.65 for agreeableness, 0.71 for
smartphone, this results in non-existent self-regulation conscientiousness, 0.86 for extraversion 0.69 for openness).
leading to addiction behaviors. Additionally, the responses were rated from 1-6 for the
Regarding the other traits, despite not having a direct Smartphone Addiction Scale-Short Version (SAS-SV)
relationship with addiction, they do influence people's where each scale shows a high level of internal consistency
behavior. For example, a high degree of extraversion (Cronbach’s  = 0.89). The data collection process lasted for
implies the need to be in contact with others. Furthermore, two weeks, aiming to obtain a representative sample from the
this trait along with a low degree of agreeableness is student population of the school.
Table 2: Description of each personality trait
Reference Personality trait Description Low trait High trait
Eichenberg et al. (2021); Lei et al. Neuroticism Index of emotional Security and trust Sensitivity and
(2020); Müller et al. (2021); stability and impulse nervousness
Peltonen et al. (2020); control
Peterka-Bonetta et al. (2019);
Toyama and Hayashi (2022);
Wickord and Quaiser-Pohl (2022)
Eichenberg et al. (2021); Erdem Conscientiousness It involves self-discipline Carelessness Efficiency and
and Uzun (2022); Müller et al. Organization and personal lack of planning organization
(2021); Peltonen et al. (2020); responsibility
Peterka-Bonetta et al. (2019);
Toyama and Hayashi (2022);
Wickord and Quaiser-Pohl (2022)
Eichenberg et al. (2021); Erdem Extraversion Characterized by energy Reserve and Expressiveness
and Uzun (2022); Peltonen et al. and taste for social preference for and search for
(2020); Toyama and Hayashi interaction solitude social interactions
(2022); Wickord and courtesy and
Quaiser-Pohl (2022) cooperation
Eichenberg et al. (2021); Erdem Agreeableness It reflects a willingness Defiance and
and Uzun (2022); Peltonen et al. to be kind, estrangement
(2020); Peterka-Bonetta et al. compassionate cooperative
(2019); Wickord and
Quaiser-Pohl (2022)
Eichenberg et al. (2021); Erdem Openness Intellectual curiosity, Caution and Imagination
and Uzun (2022); Peltonen et al. creativity is a preference conventional and curiosity
(2020); Wickord and for novelty thinking
Quaiser-Pohl (2022)

184
Jacobo Osorio et al. / Journal of Computer Science 2024, 20 (2): 181.190
DOI: 10.3844/jcssp.2024.181.190

During the data cleaning process, each survey was An analysis of the data was made and the following
carefully reviewed to ensure that there were no blank statements, which are supported by the results in Table 4,
questions. Additionally, reverse questions were included were drawn:
to ensure response validity. If a respondent answered
affirmatively to a question and then answered  The sample of students has a greater than average
affirmatively to its reverse question, it indicated degree of openness to experience and agreeableness
inconsistent responses those surveys were discarded. with an average of 34.99 and 32.88 respectively a
Within this stage, each survey was thoroughly checked standard deviation of 6.204 and 5.155 respectively.
As a result, students will have a favorable attitude
any surveys with outlier values were removed. Finally,
towards trying new experiences and will show
after the data cleaning process, a clean sample of 96
attitudes of compassion and cooperation
questionnaires was obtained from the initial total of 118
 The sample of students has a slightly higher degree of
surveys, for further analysis.
conscientiousness than the average, having a mean
To select the relevant characteristics, statistical value of 29.31 and a standard deviation of 5.935. So,
methods were employed with the main objective of students will tend to be more organized and
establishing correlations between the variables and responsible in their behaviors as well as their decisions
understanding their relationships. A total of 7 variables  The sample of students has an average degree of
were considered, as outlined in Table 3, to comprehend neuroticism and extraversion, having a mean of 23.68
the sample and examine the association between each and a standard deviation of 6.036 for neuroticism a
personality trait and the smartphone addiction intensity. mean of 26.20 with a standard deviation of 7.106 for
Utilizing the collected data and established variables, extraversion, which means that the students do not
the Pearson correlation coefficient was applied to stand out in any of those traits
analyze the influence of these variables on smartphone  The sample has a level of addiction below the average.
addiction. This analysis was conducted using the IBM Of the total, only 25% present smartphone addiction.
SPSS Statistics software, which generated the The mean value was calculated and 24.82 was
correlations depicted in (Fig. 2) and the descriptive obtained with a standard deviation of 10.154, which
statistics presented in Table 4. Considering the level of means that there is significant variability in the results,
smartphone addiction or "F06" as the primary variable, with some students presenting values considerably
a strong relationship can be observed with the higher or lower than the average, showing different
neuroticism variable or "F05" with a coefficient of levels of addiction. Thus, it can be concluded that the
0.342. This indicates a positive association between the level of addiction of the students is low
two variables, suggesting that as neuroticism increases, Table 3: Variables used for the predictive model
the likelihood of having a higher level of smartphone ID Variable Description
addiction also increases. F01 Openness Indicates the respondent's level of
On the other hand, the conscientiousness variable, openness to the experience
F02 Conscientiousness Indicates the level of awareness of the
or "F02" shows an inverse relationship with respondent
smartphone addiction, with a coefficient of -0.415. F03 Extraversion Indicates the level of extraversion of the
respondent
This implies a negative association between the two F04 Agreeableness Indicates the level of friendliness of the
variables, meaning that as conscientiousness increases, respondent
the probability of having a smartphone addiction F05 Neuroticism Indicates the level of neuroticism or
emotional instability of the respondent
decreases. Similar findings have been reported in F06 Smartphone Addiction Indicates the respondent's level of
previous research (Müller et al., 2021), where Scale (SAS) openness to the experience
F07 Age Indicates the respondent's age in years
neuroticism and conscientiousness traits have influence
coefficients of 0.379 and -0.404, respectively.
Table 4: Statistical data of the variables, obtained from IBM
However, variables such as openness (F01), SPSS Statistics
extraversion (F03) agreeableness (F04) are not closely ID Median Description No.
related to smartphone addiction; they present coefficients F01 34.99 6,204 96
of -0.082, and -0.079 0.039 respectively. These values F02 29.31 5,935 96
F03 26.20 7,106 96
indicate a weak or insignificant influence on smartphone
F04 32.88 5,155 96
addiction based on the sample. As in (Müller et al., 2021), F05 23.68 6,036 96
these traits influence people's behavior, but they are not F06 24.82 10,154 96
relevant to smartphone addiction. F07 14.01 1,244 96

185
Jacobo Osorio et al. / Journal of Computer Science 2024, 20 (2): 181.190
DOI: 10.3844/jcssp.2024.181.190

algorithms using nodes from the platform as well as

performing automated training for machine learning,
which facilitated the training process and optimized the
performance of the models.
The collected data was uploaded to the platform and
the "select columns in dataset" component was used to
define the relevant columns for the algorithm training,
where age and the results of each personality trait in the
SAS questionnaire were considered. The individual
answers to each question were not considered to simplify
algorithm training and to reduce the number of required
parameters, which gave us greater interpretability.
Subsequently, a "split data" component was used to
establish that 70% of the data was used for training and
the remaining 30% for testing, in other words, 67 surveys
were used for training and 29 surveys for testing. These
samples were used to discover if it was possible to
determine the respondents' addiction based on their results
from the questionnaires.
Figure 3 illustrates the components utilized in the
model. The "train model" component was employed for
Fig. 2: Correlation between the variables and the SAS-SV
training purposes, while the "score model" component
was used for conducting tests and making predictions.
Evaluation
Finally, the "evaluate model" component was utilized
to analyze the algorithm's results. The same configuration
was applied to both the DT and LR algorithms. In contrast
to the three previous algorithms, in the case of the XGB
algorithm, it was decided to use the "automated ML"
procedure because the platform did not have the
corresponding nodes for training.
The performance of the algorithms was evaluated
using specific metrics (Jiménez et al., 2023), each with its
own corresponding (Eqs. 1-5):
𝑇𝑃 + 𝑇𝑁
𝐴𝑐𝑐𝑢𝑟𝑎𝑐𝑦 = (1)
𝑇𝑃 + 𝑇𝑁 + 𝐹𝑃 + 𝐹𝑁

𝑇𝑃
𝑃𝑟𝑒𝑐𝑖𝑠𝑖𝑜𝑛 = (2)
𝑇𝑃 + 𝐹𝑃

𝑇𝑃
𝑆𝑒𝑛𝑠𝑖𝑡𝑖𝑣𝑖𝑡𝑦 = (3)
𝑇𝑃 + 𝐹𝑁

𝑇𝑁
𝑆𝑝𝑒𝑐𝑖𝑓𝑖𝑐𝑖𝑡𝑦 = (4)
𝑇𝑁 + 𝐹𝑃

Fig. 3: Components used for random forest training 𝐹𝑎𝑙𝑙 𝑂𝑢𝑡 =

𝐹𝑃
(5)
𝐹𝑃 + 𝑇𝑁

Modeling
Results and Discussion
This phase describes the classification techniques
that will be used in this study. The modeling After the training, the Confusion Matrix that evaluated
development was conducted using the Azure Machine the performance of each one of the algorithms was obtained.
Learning Studio platform. Throughout the process, a This can be seen in Fig. 4, generally giving us favorable and
compute instance was employed to execute the Python similar results for all the algorithms, highlighting the RF
code effectively. Furthermore, a cluster specifically algorithm (Fig. 4a), which obtained more correct predictions
designed for machine learning tasks was created as an than incorrect ones. The interpretation of the variables used
add-on. These instances had the objective of training the for the metrics can be seen in Table 5.

186
Jacobo Osorio et al. / Journal of Computer Science 2024, 20 (2): 181.190
DOI: 10.3844/jcssp.2024.181.190

Pridicted Label The RF algorithm (Fig. 4a), XGB algorithm (Fig. 4b)
No SI DT algorithm (Fig. 4c) demonstrated a higher number of
correct predictions for true negatives, indicating their
proficiency in identifying cases where addiction is not
No

present. On the other hand, the LR algorithm (Fig. 4d)

yielded more true positives. However, it also exhibited a
True Label

higher number of false positives, suggesting a tendency to

misclassify certain cases.
The results presented in Table 6 emphasize the
SI

performance of each algorithm in accurately predicting

and minimizing errors.
(a) Using the Azure Machine Learning Studio platform, the
algorithms' performance was evaluated by calculating
Pridicted Label
various metrics, providing a comprehensive perspective of
No SI the models' performance. The results, as shown in Table 7,
offer insights into the algorithms' performance. It is worth
mentioning that the results obtained through automated
calculations may have slight variations compared to
No

manually computed values from the confusion matrix.

True Label

The tree-based algorithms (RF, XGB DT) demonstrated

higher accuracy, ranging from 83-87%, compared to LR.
However, LR achieved better results in the sensitivity
metric, surpassing 70%. Despite the similarity in results,
SI

the RF algorithm outperformed the others in this study,

with a precision of 87.3% and an accuracy of 89.7%. These
metrics indicate its lower likelihood of false positives and
(b)
its overall ability to accurately classify smartphone
addiction based on personality traits.
Pridicted Label
No SI
Table 5: Interpretation of the variables used for the confusion matrix
Code Variable Description
TP True Positive Correct predictions of adolescents with
smartphone addiction based on their
personality traits
No

TN True Negative Correct predictions of adolescents without

True Label

smartphone addiction based on their

personality traits
FP False Positive Incorrect predictions of adolescents with
smartphone addiction based on their
SI

personality traits
FN False Negative Incorrect predictions of adolescents without
smartphone addiction based on their
(c) personality traits

Table 6: Results of the number of correct and incorrect predictions

Pridicted Label according to the algorithm
No SI
Correct Incorrect
Algorithm Result predictions predictions
Random forest With addiction 5 1
Without addiction 21 2
No

XG boost With addiction 4 1

True Label

Without addiction 21 3
Decision tree With addiction 4 1
Without addiction 21 3
Logistic regression With addiction 4 1
Without addiction 21 3
SI

Table 7: Comparative results of performance metrics of trained algorithms

Algorithm Accuracy Precision Sensitivity Fall Out Specificity AUC
(d) Random forest 0.897 0.873 0.669 0.045 0.955 0.987
XG boost 0.862 0.838 0.526 0.045 0.955 0.948
Decision tree 0.862 0.857 0.526 0.045 0.955 0.938
Fig. 4: Confusion matrix of trained algorithms Logistic regression 0.862 0.808 0.720 0.136 0.864 0.942

187
Jacobo Osorio et al. / Journal of Computer Science 2024, 20 (2): 181.190
DOI: 10.3844/jcssp.2024.181.190

exceeding 90%. Notably, the ROC curve of the RF

algorithm (Fig. 5a) demonstrates a closer proximity to 1
on the Y-axis, indicating a higher overall accuracy in
distinguishing between positive and negative instances. It
also demonstrates superior performance with a high ROC
curve, signifying a greater probability of accurately
classifying positive instances. This is supported by Table 7,
which shows the highest AUC value of 98.7%.
When considering the influence of personality traits,
the RF algorithm also exhibits the highest predictive
(a) performance for addiction. It is remarkable precision and
sensitivity values validate its ability to effectively identify
the relationship between personality traits and smartphone
addiction. In contrast, the XGB (Fig. 5b), DT (Fig. 5c) LR
(Fig. 5d) algorithms demonstrate lower precision and
recall, indicating comparatively inferior performance.
The complexity of the RF algorithm, which arises
from the ensemble nature of decision trees, was
effectively managed through parameter tuning and feature
selection techniques. This allowed us to strike a balance
between model complexity and interpretability. We found
that by carefully selecting hyperparameters and leveraging
feature importance scores, we could maintain a reasonable
(b) level of model transparency while still preserving the
algorithm's exceptional predictive capability.

Conclusion
Based on the proposed model using the RF algorithm,
there is evidence of a relationship between personality
traits and smartphone addiction in teenagers, highlighting
the traits of neuroticism and conscientiousness. It is worth
mentioning that, as of now no study has aimed to predict
smartphone addiction based on personality traits.
Based on Pearson's correlation analysis, it was
(c) identified that the variables showing the strongest
correlation with addiction are "Conscientiousness" (F02),
"Neuroticism" (F05) "Age" (F07). Additionally, an
inverse relationship was found between "Neuroticism"
(F05) and all other personality traits, except for "Openness
to experience" (F01). This finding substantiates previous
research results, which have consistently shown similar
outcomes for neuroticism and conscientiousness traits in
relation to addiction (Müller et al., 2021).
Among the trained algorithms, the RF algorithm
demonstrated superior suitability with an accuracy of
89.7% and a precision of 87.3%. It showed accurate
predictions and high overall performance, achieving the
(d)
highest AUC value in the ROC curve, which indicates
Fig. 5: ROC curve of the trained algorithms better classification probability for positive instances.
In future work, it is recommended to incorporate new
To assess the performance of the classification data to enhance the performance of the algorithms used. It
models, the Area Under the Curve (AUC) metric was is also important to evaluate the algorithms' performance
utilized, which offered insights into the relationship with external data for cross-validation to ensure the
between "Fall out" and sensitivity. Figure 5 illustrates the robustness of the results. Finally, exploring other relevant
ROC curves for each algorithm, all exhibiting values variables that influence addiction prediction, such as the

188
Jacobo Osorio et al. / Journal of Computer Science 2024, 20 (2): 181.190
DOI: 10.3844/jcssp.2024.181.190

environment, social conditions, or economic resources Chen, W. Y., Yan, L., Yuan, Y. R., Zhu, X. W., Zhang,
could be very beneficial. Y. H., & Lian, S. L. (2021). Preference for solitude
and Mobile phone addiction among Chinese college
Acknowledgment students: the mediating role of psychological distress
and moderating role of mindfulness. Frontiers in
We are deeply grateful to the university, the school the
Psychology, 12, 750511.
students who conscientiously participated in our research https://doi.org/10.3389/fpsyg.2021.750511
by completing the surveys and thereby contributing to our Chen, Y., Zhang, X., Lu, L., Wang, Y., Liu, J., Qin, L., ...
database. Our appreciation also extends to the & Chen, M. C. (2022). Machine learning methods to
psychologist who played a crucial role in developing and identify predictors of psychological distress.
validating the survey. Processes, 10(5), 1030.
https://doi.org/10.3390/PR10051030
Funding Information Cheng, Y., & Meng, J. (2021). The association between
This research was financed by the research Department depression and problematic smartphone behaviors
of the Universidad Peruana de Ciencias Aplicadas. through smartphone use in a clinical sample. Human
Behavior and Emerging Technologies, 3(3), 441-453.
https://doi.org/10.1002/HBE2.258
Author’s Contributions
Duan, L., He, J., Li, M., Dai, J., Zhou, Y., Lai, F., & Zhu,
Jacobo Osorio and Marko Figueroa: Literature analysis, G. (2021). Based on a decision tree model for
data collection, model and implementation of models, exploring the risk factors of smartphone addiction
experimentation analysis of results. Manuscript written. among children and adolescents in China during the
Lenis Wong: Study supervision, result analysis, COVID-19 pandemic. Frontiers in Psychiatry, 12,
manuscript reviewed discussion. 652356. https://doi.org/10.3389/fpsyt.2021.652356
Eichenberg, C., Schott, M., & Schroiff, A. (2021).
Ethics Problematic smartphone use-comparison of students
The article is authentic and contains unpublished with and without problematic smartphone use in light
material. The corresponding author affirms that no ethical of personality. Frontiers in Psychiatry, 11, 599241.
concerns exist all authors have read and endorsed the article. https://doi.org/10.3389/fpsyt.2020.599241
Erdem, C., & Uzun, A. M. (2022). Smartphone addiction
among undergraduates: Roles of personality traits
References
and demographic factors. Technology, Knowledge
Abu-Taieh, E. M., AlHadid, I., Kaabneh, K., and Learning, 27(2), 579-597.
Alkhawaldeh, R. S., Khwaldeh, S., Masa’deh, R. E., https://doi.org/10.1007/s10758-020-09467-1
& Alrowwad, A. A. (2022). Predictors of Smartphone Garriga, R., Mas, J., Abraha, S., Nolan, J., Harrison, O.,
Addiction and Social Isolation among Jordanian Tadros, G., & Matic, A. (2022). Machine learning
Children and Adolescents Using SEM and ML. Big model to predict mental health crises from electronic
Data and Cognitive Computing, 6(3), 92. health records. Nature Medicine, 28(6), 1240-1248.
https://doi.org/10.3390/BDCC6030092 https://doi.org/10.1038/s41591-022-01811-5
Blasi, A. H., & Alsuwaiket, M. (2020). Analysis of Haque, U. M., Kabir, E., & Khanam, R. (2021). Detection
students' misconducts in higher education using of child depression using machine learning methods.
decision tree and ann algorithms. Engineering, PLoS One, 16(12), e0261131.
Technology & Applied Science Research, 10(6), https://doi.org/10.1371/JOURNAL.PONE.0261131
6510-6514. https://doi.org/10.48084/etasr.3927
Castillo-Viera, E., Gago-Valiente, F.-J., Giménez- Ipsos. (2021). El Internet de Cada Día: Hábitos y
Fuentes-Guerra, F.-J., Abad-Robles, M.-T., & Actitudes Hacia El Internet.
Moreno-Sánchez, E. (2022). Physical Activity https://www.ipsos.com/es-pe/el-internet-de-cada-
Programmes in the Treatment of Addictions: A dia-habitos-y-actitudes-hacia-el-internet-2021
Systematic Review. Applied Sciences, 12(18), 9117. Jiménez, O., Jesús, A., & Wong, L. (2023, May). Model for
https://doi.org/10.3390/app12189117 the Prediction of Dropout in Higher Education in Peru
Cha, S. S., & Seo, B. K. (2018). Smartphone use and applying Machine Learning Algorithms: Random
smartphone addiction in middle school students in Forest, Decision Tree, Neural Network and Support
Korea: Prevalence, social networking service and Vector Machine. In 2023 33rd Conference of Open
game use. Health Psychology Open, 5(1), Innovations Association (FRUCT) (pp. 116-124). IEEE.
2055102918755046. https://doi.org/10.23919/FRUCT58615.2023.10143
https://doi.org/10.1177/2055102918755046 068

189
Jacobo Osorio et al. / Journal of Computer Science 2024, 20 (2): 181.190
DOI: 10.3844/jcssp.2024.181.190

Kim, S., Lee, H. K., & Lee, K. (2021). Screening of mood Nida. (2022). La Adicción y La Salud.
symptoms using MMPI-2-RF scales: An application https://nida.nih.gov/es/publicaciones/las-drogas-el-
of machine learning techniques. Journal of cerebro-y-la-conducta-la-ciencia-de-la-adiccion/la-
Personalized Medicine, 11(8), 812. adiccion-y-la-salud
https://doi.org/10.3390/JPM11080812 Peltonen, E., Sharmila, P., Asare, K. O., Visuri, A.,
Lee, J., & Kim, W. (2021). Prediction of problematic Lagerspetz, E., & Ferreira, D. (2020). When phones
smartphone use: A machine learning approach. get personal: Predicting Big Five personality traits
International Journal of Environmental Research from application usage. Pervasive and Mobile
and Public Health, 18(12), 6458. Computing, 69, 101269.
https://doi.org/10.3390/ijerph18126458 https://doi.org/10.1016/J.PMCJ.2020.101269
Lei, L. Y. C., Ismail, M. A. A., Mohammad, J. A. M., & Peralta, F. C. (2014, October). Fases del modelo de
Yusoff, M. S. B. (2020). The relationship of proceso de la metodología CRISP-DM.
smartphone addiction with psychological distress and https://www.researchgate.net/figure/Fases-del-
neuroticism among university medical students. modelo-de-proceso-de-la-metodologia-CRISP-DM-
BMC Psychology, 8, 1-9. 35_fig2_284215308
https://doi.org/10.1186/S40359-020-00466-6 Peterka-Bonetta, J., Sindermann, C., Elhai, J. D., &
Li, X., Fu, S., Fu, Q., & Zhong, B. (2021). Youths’ Montag, C. (2019). Personality associations with
habitual use of smartphones alters sleep quality and smartphone and internet use disorder: A comparison
memory: Insights from a national sample of Chinese study including links to impulsivity and social
students. International Journal of Environmental anxiety. Frontiers in Public Health, 7, 127.
Research and Public Health, 18(5), 2254. https://doi.org/10.3389/FPUBH.2019.00127
https://doi.org/10.3390/ijerph18052254 Razavi, R., Gharipour, A., & Gharipour, M. (2020).
Li, Y., Wang, Z., You, W., & Liu, X. (2022). Core self- Depression screening using mobile phone usage
evaluation, mental health and mobile phone dependence metadata: A machine learning approach. Journal of
in Chinese high school students: Why should we care. the American Medical Informatics Association,
Italian Journal of Pediatrics, 48(1), 28. 27(4), 522-530.
https://doi.org/10.1186/S13052-022-01217-6 https://doi.org/10.1093/jamia/OCZ221
Makino, K., Lee, S., Bae, S., Chiba, I., Harada, K., Rho, M. J., Park, J., Na, E., Jeong, J. E., Kim, J. K., Kim,
Katayama, O., ... & Shimada, H. (2021). D. J., & Choi, I. Y. (2019). Types of problematic
Development and validation of new screening tool smartphone use based on psychiatric symptoms.
for predicting dementia risk in community-dwelling Psychiatry Research, 275, 46-52.
older Japanese adults. Journal of Translational https://doi.org/10.1016/J.PSYCHRES.2019.02.071
Medicine, 19, 1-9. https://doi.org/10.1186/S12967- Toyama, M., & Hayashi, Y. (2022). Links of personality
021-03121-9 traits to media multitasking: Conscientiousness
Marengo, D., Sindermann, C., Häckel, D., Settanni, M., predicts mobile phone use in the college classroom.
Elhai, J. D., & Montag, C. (2020). The association Current Psychology, 41(12), 8660-8667.
https://doi.org/10.1007/S12144-020-01258-2
between the Big Five personality traits and
Wickord, L. C., & Quaiser-Pohl, C. (2022).
smartphone use disorder: A meta-analysis. Journal of
Psychopathological symptoms and personality traits
Behavioral Addictions, 9(3), 534-550.
as predictors of problematic smartphone use in
https://doi.org/10.1556/2006.2020.00069 different age groups. Behavioral Sciences, 12(2), 20.
Minsa. (2021). La Pandemia COVID-19 Ha Incrementado https://doi.org/10.3390/bs12020020
La Adicción a Los Videojuegos En Niños y Xia, F., Li, Q., Luo, X., & Wu, J. (2022). Machine
Adolescentes. learning model for depression based on heavy metals
https://www.gob.pe/institucion/minsa/noticias/3480 among aging people: A study with National Health
05-minsa-la-pandemia-covid-19-ha-incrementado- and Nutrition Examination Survey 2017-2018.
la-adiccion-a-los-videojuegos-en-ninos-y- Frontiers in Public Health, 10, 939758.
adolescentes https://doi.org/10.3389/FPUBH.2022.939758
Müller, M., Sindermann, C., Rozgonjuk, D., & Montag, Zeighami, R., Hosseini, S. M., Mohammadi, N., &
C. (2021). Mind-wandering mediates the associations Shahsavari, S. (2021). Predicting addiction
between neuroticism and conscientiousness, and susceptibility based on big five personality traits.
tendencies towards smartphone use disorder. Iranian Journal of Psychiatry and Behavioral
Frontiers in Psychology, 12, 661541. Sciences, 15(4).
https://doi.org/10.3389/fpsyg.2021.661541 https://doi.org/10.5812/IJPBS.109103

190